3-D object reconstruction from multiple 2-D images

AbstractThree-dimensional (3-D) object reconstruction from multiple two-dimensional (2-D) images is one of the most important topics in computer vision. In this paper, we review two common methods for 3-D object reconstruction: shape from silhouette and voxel coloring. The shape from silhouette method recovers the 3-D shape of an object from silhouette images. This method is based on the visual hull that is formed by intersecting volumes from silhouette cones. Voxel coloring is another reconstruction method that uses the measure of color consistency to build a 3-D model of the object. In this paper, we present a different 3-D object reconstruction algorithm, where the initial 3-D object that was generated by the visual hull is carved to represent details of the 3-D object using color consistency. In this algorithm, visibility checking is proposed to reconstruct the 3-D object well. Experimental results show that the proposed algorithm reconstructs natural 3-D models efficiently.